Evaluating geometric accuracy of multi-platform stereotactic neuroimaging in radiosurgery

We used a spherical phantom to evaluate geometric accuracy in multi-platform stereotactic neuroimaging for radiosurgery. The phantom consisted of two plastic 16-cm-diameter hemispheres in which an exchangeable 8-cm plastic functional cube was incorporated. The functional cube contained cylinder and point targets. The targets were filled with a mixed aqueous solution of 2-mM copper sulfate and 300-mg/ml iodinated contrast medium and were visible on both MR and X-ray images. Two MR scanners and a biplane X-ray angio-suite were used to scan the phantom stereotactically in two sessions of the experiment. The angio-suite was equipped with digital subtraction and distortion-correction software. The resulting stereotactic images were transferred to a dose-planning computer for length measurement and coordinate determination of the targets. The mean errors of the measured cylinder length on distortion non-corrected X-ray stereotactic images were 0.24 +/- 0.14 and 0.73 +/- 0.10 mm, respectively, in the experiments; on distortion-corrected images 0.22 +/- 0.10 and 0.35 +/- 0.39 mm. They were 0.50 +/- 0.24, 0.25 +/- 0.19 and 0.49 +/- 0.34, 0.23 +/- 0.25 mm, respectively, of the two MR scanners. The mean errors of coordinate determination of point targets between the stereotactic MR and the distortion-corrected X-ray images were 0.70 +/- 0.18, 0.52 +/- 0.22 and 0.76 +/- 0.25, 0.40 +/- 0.10 mm, respectively, in the experiments. We found that the overall geometric errors of target delineation between stereotactic MR and X-ray images were in the submillimeter range. The current study validates the multi-platform and multi-facility stereotactic neuroimaging practice and ensures imaging accuracy in radiosurgery.